To add to Jonjova's answer:
for each [i,j], you would execute the following call:
double noiseValue = SimplexNoise.noise(i * widthFactor, j * heightFactor);
This will return values that range from -1 to 1. How you convert those values to ones that are useful for your voxels is up to you. The two most common strategies are scaling ("smooth" noise) and applying an Abs function to the results prior to scaling ("turbulent" noise).
The values widthFactor and hightFactor above are used to control how quickly the values transition. With SimplexNoise, progressing integers tend to just generate unrelated, random values. I'd try using a factor such as 1/128 (give or take a power of two) as then you will get the transitions between the random peaks.
To get more interesting effects, you will have to make multiple calls to the function at differing width/height factors, and weight the various answers depending on the frequency content you desire. Also, the noise result can be used to permute another function such as a sin or radial gradient.